旋转机械升降速过程的双谱-FHMM识别方法

Study on Bispectrum-FHMM Recognition Method in Speed-up and Speed-down Process of Rotating Machinery

结合双谱和因子隐 Markov模型 ,提出了一种基于双谱的特征提取建立机组各状态相应的因子隐 Markov模型状态识别法 ,并成功地应用到旋转机械升降速过程的故障诊断中 ,同时还与基于双谱的特征提取的 HMM状态识别法进行了比较 ,实验结果表明该方法是有效的。

Bispectrum is a useful tool for processing non-Gaussian signal and nonlinear system. Factorial hidden Markov models (FHMM), which is a generalization of HMM, is superior to HMM, and has a capability of pattern recognitaion baseded on time series, particularly suitable for signal which is non-stationary, bad repetition and reappearance. An approach of fault diagnosis using speed-up and spped-down process of rotating machinery, combining bispectrum with FHMM, is proposed, which is that bispectrum is used as a fault feature, and FHMM as a classifier. This approach is compared with another classfication approach in which bispectrum is used as a fault feature, HMM as a classifier. Experiment results show that this approach is very effective.